AIMS/HYPOTHESIS:The calibre of the retinal vessels has been linked to diabetes mellitus but studies have not shown consistent results. We conducted a participant-level meta-analysis to evaluate the association between retinal arteriolar and venular calibre and diabetes. METHODS:We performed a systematic review on MEDLINE and EMBASE for articles published up to December 2014. We identified five population-based prospective cohort studies that provided individual-level data on 18,771 diabetes-free participants. We used discrete time proportional hazards models to estimate pooled HRs of diabetes associated with 1 SD (20 μm) change in retinal vascular calibre. RESULTS:We identified 2,581 incident cases of diabetes over a median follow-up period of 10 years (interquartile interval of 3.4-15.8 years). After adjustment for demographic, lifestyle and clinical factors, retinal venular calibre was significantly associated with incident diabetes (pooled HR 1.09 [95% CI 1.02, 1.15] per SD increase in venular calibre). This association persisted in analyses excluding individuals with <5 years of follow-up (1.07 [1.0, 1.12]) or those with impaired fasting glucose at baseline (1.10 [1.03, 1.17]); in subgroup analyses, the association was stronger in men than in women but was consistent across subgroups of race/ethnicity, smoking status, hypertension and BMI categories. Retinal arteriolar calibre was not associated with diabetes (0.95 [0.86, 1.06] per SD decrease in arteriolar calibre). CONCLUSIONS/INTERPRETATION:Wider retinal venules but not narrower retinal arterioles were associated with a modestly increased risk for diabetes. Knowledge of pathological mechanisms underlying wider retinal venule may provide further insights concerning microvascular alterations in diabetes.